[llvm] e8be733 - [VPlan] Remove redundant ExtractLastElement from vector-to-scalar VPI.

[Florian Hahn via llvm-commits]

Author: Florian Hahn
Date: 2025-06-20T12:45:20+01:00
New Revision: e8be733a3c3347207c162fc83e8dbe02dad2a952

URL: https://github.com/llvm/llvm-project/commit/e8be733a3c3347207c162fc83e8dbe02dad2a952
DIFF: https://github.com/llvm/llvm-project/commit/e8be733a3c3347207c162fc83e8dbe02dad2a952.diff

LOG: [VPlan] Remove redundant ExtractLastElement from vector-to-scalar VPI.

Recipes that are vector-to-scalar are guaranteed to generate a scalar
value, so the extract is redundant after VPlan unrolling. Remove it.

This removes unneeded ExtractLastElement VPInstruction of reduction
result computations.

Added: 
    

Modified: 
    llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
    llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
    llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index b2535fe3aa578..ae91e310f7592 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -192,6 +192,9 @@ inline match_combine_and<LTy, RTy> m_CombineAnd(const LTy &L, const RTy &R) {
 /// Match a VPValue, capturing it if we match.
 inline bind_ty<VPValue> m_VPValue(VPValue *&V) { return V; }
 
+/// Match a VPInstruction, capturing if we match.
+inline bind_ty<VPInstruction> m_VPInstruction(VPInstruction *&V) { return V; }
+
 template <typename Ops_t, unsigned Opcode, bool Commutative,
           typename... RecipeTys>
 struct Recipe_match {

diff  --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 60a8837fb76aa..b6ca50549fa35 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1163,6 +1163,14 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
     });
     return;
   }
+
+  VPInstruction *OpVPI;
+  if (match(Def, m_VPInstruction<VPInstruction::ExtractLastElement>(
+                     m_VPInstruction(OpVPI))) &&
+      OpVPI->isVectorToScalar()) {
+    Def->replaceAllUsesWith(OpVPI);
+    return;
+  }
 }
 
 void VPlanTransforms::simplifyRecipes(VPlan &Plan, Type &CanonicalIVTy) {

diff  --git a/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll b/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
index 8095f258ea183..307ab3a87e9d9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
@@ -106,13 +106,12 @@ define i32 @print_partial_reduction(ptr %a, ptr %b) {
 ; CHECK-EMPTY:
 ; CHECK-NEXT: middle.block:
 ; CHECK-NEXT:   EMIT vp<[[RED_RESULT:%.+]]> = compute-reduction-result ir<%accum>, ir<%add>
-; CHECK-NEXT:   EMIT vp<[[EXTRACT:%.+]]> = extract-last-element vp<[[RED_RESULT]]>
 ; CHECK-NEXT:   EMIT vp<[[CMP:%.+]]> = icmp eq ir<1024>, ir<1024>
 ; CHECK-NEXT:   EMIT branch-on-cond vp<[[CMP]]>
 ; CHECK-NEXT: Successor(s): ir-bb<exit>, ir-bb<scalar.ph>
 ; CHECK-EMPTY:
 ; CHECK-NEXT: ir-bb<exit>:
-; CHECK-NEXT:   IR   %add.lcssa = phi i32 [ %add, %for.body ] (extra operand: vp<[[EXTRACT]]> from middle.block)
+; CHECK-NEXT:   IR   %add.lcssa = phi i32 [ %add, %for.body ] (extra operand: vp<[[RED_RESULT]]> from middle.block)
 ; CHECK-NEXT: No successors
 ; CHECK-EMPTY:
 ; CHECK-NEXT: ir-bb<scalar.ph>: